Modelling the brain response to arbitrary visual stimulation patterns for a flexible high-speed Brain-Computer Interface

<b>1. General description</b>Data was recorded using BCI2000 with g.USBamp (g.tec, Austria) EEG amplifier. 32 electrodes were used. Sampling rate was set to 600 Hz and data was bandpass filtered by the amplifier between 0.1 Hz and 60 Hz using a Chebyshev filter of order 8 and notch-filtered at 50 Hz. Data was stored as MATLAB mat-File, whereas each file is the data of one participant. <br><b>2. Experimental description</b>The experiment was split in three parts: spatial filter runs, training runs, testing runs. The stimulation patterns were presented with 60 bits per second and each run consits of 32 trials. During the testing runs, the runs were alternated between using random stimulation patterns and optimized stimulation patterns. As all trials were concatenated, this has to be taken into account. Furthermore, the participants had to perform the trials of each run in lexicographic order.<b><br></b><b>3. Variable description</b>Each file VP*.mat contains the following variables:- spfilter_data_x contains the raw EEG data of the spatial filter runs split by each m-sequence cyle. As there were 3 runs with 32 trials and 3 m-sequence cycles each, there are 3*32*3=192 m-sequence cycles in total. Since the m-sequence was shifted by 2 bits for each succesive target, the trials were lag-fixed previously. As the m-sequence consists of 63 bits, the data consists of 63/60*600=630 samples. Therefore the matrix has the following dimension: #m-sequences X #channels X #samples- spfilter_data_y contains the stimulation pattern for each m-sequenze cycle upsampled to be synchronized with the EEG data. The matrix has the following dimension: #m-sequences X #samples- train_data_x_*s contains the raw EEG data of the training runs split by trials. As different trial durations were used during training, * denotes the duration in seconds. The matrix has the following dimension: #trials X #channels X #samples- train_data_y_*s contains the stimulation pattern for each target during the training runs, upsampled to be synchronized with the EEG data. As the layout consits of 32 targets, the matrix has the following dimension: #trials X #targets X #samples- test_data_x contains the raw EEG data of the training runs split by trials. The trial duration was 2 seconds, therefore, there are 1200 samples per trial. The matrix has the following dimension: #trials X #channels X #samples- test_data_y contains the stimulation pattern for each target during the testing runs, upsampled to be synchronized with the EEG data. As the layout consits of 32 targets, the matrix has the following dimension: #trials X #targets X #samples There is an additional file "targetdelys.mat" containing a variable with the number of samples for each target (in lexicographic order) used to correct the raster latency.

1. 总体概述
本数据集采用搭载奥地利g.tec公司g.USBamp脑电（Electroencephalogram, EEG）放大器的BCI2000系统采集，共使用32通道电极。采样率设置为600Hz，数据经放大器采用8阶切比雪夫滤波器（Chebyshev Filter）完成0.1Hz~60Hz带通滤波，并在50Hz处进行陷波滤波。数据以MATLAB的.mat格式存储，每个文件对应一名被试的采集数据。
2. 实验流程说明
本实验分为三个阶段：空间滤波任务阶段、训练任务阶段与测试任务阶段。刺激模式以60比特每秒的速率呈现，每个任务阶段包含32次试次。在测试任务阶段，实验会交替使用随机刺激模式与优化后的刺激模式。由于所有试次均为拼接得到，处理数据时需注意这一特性。此外，被试需按照字典序完成每个任务阶段内的所有试次。
3. 变量说明
每个VP*.mat格式文件包含以下变量：
- spfilter_data_x：存储按每个m序列（m-sequence）周期拆分的空间滤波任务阶段原始脑电数据。本实验共包含3个任务阶段，每个阶段含32次试次，且每个阶段对应3个m序列周期，因此总共有3×32×3=192个m序列周期。由于每个后续目标对应的m序列会偏移2比特，试次已预先完成滞后校正。单个m序列包含63比特，因此每个周期对应63/60×600=630个采样点。该矩阵的维度为：#m序列周期数 × #通道数 × #采样点数
- spfilter_data_y：存储每个m序列周期对应的刺激模式，该模式已进行上采样以与脑电数据同步。该矩阵的维度为：#m序列周期数 × #采样点数
- train_data_x_*s：存储按试次拆分的训练任务阶段原始脑电数据。由于训练阶段采用了不同时长的试次，*代表试次时长（单位：秒）。该矩阵的维度为：#试次数 × #通道数 × #采样点数
- train_data_y_*s：存储训练阶段中每个目标对应的刺激模式，该模式已进行上采样以与脑电数据同步。由于刺激布局包含32个目标，该矩阵的维度为：#试次数 × #目标数 × #采样点数
- test_data_x：存储按试次拆分的训练任务阶段原始脑电数据。本次实验中试次时长为2秒，因此每个试次包含1200个采样点。该矩阵的维度为：#试次数 × #通道数 × #采样点数
- test_data_y：存储测试阶段中每个目标对应的刺激模式，该模式已进行上采样以与脑电数据同步。由于刺激布局包含32个目标，该矩阵的维度为：#试次数 × #目标数 × #采样点数
此外，还有一个额外文件"targetdelys.mat"，其中包含一个变量，存储按字典序排列的每个目标对应的采样点数，用于校正光栅延迟。